cuda_toolkit/include/thrust/system/tbb/detail/merge.inl

/*
 *  Copyright 2008-2021 NVIDIA Corporation
 *
 *  Licensed under the Apache License, Version 2.0 (the "License");
 *  you may not use this file except in compliance with the License.
 *  You may obtain a copy of the License at
 *
 *      http://www.apache.org/licenses/LICENSE-2.0
 *
 *  Unless required by applicable law or agreed to in writing, software
 *  distributed under the License is distributed on an "AS IS" BASIS,
 *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 *  See the License for the specific language governing permissions and
 *  limitations under the License.
 */

#pragma once

#include <thrust/detail/config.h>

#if defined(_CCCL_IMPLICIT_SYSTEM_HEADER_GCC)
#  pragma GCC system_header
#elif defined(_CCCL_IMPLICIT_SYSTEM_HEADER_CLANG)
#  pragma clang system_header
#elif defined(_CCCL_IMPLICIT_SYSTEM_HEADER_MSVC)
#  pragma system_header
#endif // no system header

#include <thrust/iterator/iterator_traits.h>
#include <thrust/detail/temporary_array.h>
#include <thrust/system/tbb/detail/execution_policy.h>
#include <thrust/merge.h>
#include <thrust/binary_search.h>
#include <thrust/detail/seq.h>
#include <tbb/parallel_for.h>

THRUST_NAMESPACE_BEGIN
namespace system
{
namespace tbb
{
namespace detail
{
namespace merge_detail
{

template<typename InputIterator1,
         typename InputIterator2,
         typename OutputIterator,
         typename StrictWeakOrdering>
struct range
{
  InputIterator1 first1, last1;
  InputIterator2 first2, last2;
  OutputIterator result;
  StrictWeakOrdering comp;
  size_t grain_size;

  range(InputIterator1 first1, InputIterator1 last1,
        InputIterator2 first2, InputIterator2 last2,
        OutputIterator result,
        StrictWeakOrdering comp,
        size_t grain_size = 1024)
    : first1(first1), last1(last1),
      first2(first2), last2(last2),
      result(result), comp(comp), grain_size(grain_size)
  {}

  range(range& r, ::tbb::split)
    : first1(r.first1), last1(r.last1),
      first2(r.first2), last2(r.last2),
      result(r.result), comp(r.comp), grain_size(r.grain_size)
  {
    // we can assume n1 and n2 are not both 0
    size_t n1 = thrust::distance(first1, last1);
    size_t n2 = thrust::distance(first2, last2);

    InputIterator1 mid1 = first1;
    InputIterator2 mid2 = first2;

    if (n1 > n2)
    {
      mid1 += n1 / 2;
      mid2 = thrust::lower_bound(thrust::seq, first2, last2, raw_reference_cast(*mid1), comp);
    }
    else
    {
      mid2 += n2 / 2;
      mid1 = thrust::upper_bound(thrust::seq, first1, last1, raw_reference_cast(*mid2), comp);
    }

    // set first range to [first1, mid1), [first2, mid2), result
    r.last1 = mid1;
    r.last2 = mid2;

    // set second range to [mid1, last1), [mid2, last2), result + (mid1 - first1) + (mid2 - first2)
    first1 = mid1;
    first2 = mid2;
    result += thrust::distance(r.first1, mid1) + thrust::distance(r.first2, mid2);
  }

  bool empty(void) const
  {
    return (first1 == last1) && (first2 == last2);
  }

  bool is_divisible(void) const
  {
    return static_cast<size_t>(thrust::distance(first1, last1) + thrust::distance(first2, last2)) > grain_size;
  }
};

struct body
{
  template <typename Range>
  void operator()(Range& r) const
  {
    thrust::merge(thrust::seq,
                  r.first1, r.last1,
                  r.first2, r.last2,
                  r.result,
                  r.comp);
  }
};

} // end namespace merge_detail

namespace merge_by_key_detail
{

template<typename InputIterator1,
         typename InputIterator2,
         typename InputIterator3,
         typename InputIterator4,
         typename OutputIterator1,
         typename OutputIterator2,
         typename StrictWeakOrdering>
struct range
{
  InputIterator1 keys_first1, keys_last1;
  InputIterator2 keys_first2, keys_last2;
  InputIterator3 values_first1;
  InputIterator4 values_first2;
  OutputIterator1 keys_result;
  OutputIterator2 values_result;
  StrictWeakOrdering comp;
  size_t grain_size;

  range(InputIterator1 keys_first1, InputIterator1 keys_last1,
        InputIterator2 keys_first2, InputIterator2 keys_last2,
        InputIterator3 values_first1,
        InputIterator4 values_first2,
        OutputIterator1 keys_result,
        OutputIterator2 values_result,
        StrictWeakOrdering comp,
        size_t grain_size = 1024)
    : keys_first1(keys_first1), keys_last1(keys_last1),
      keys_first2(keys_first2), keys_last2(keys_last2),
      values_first1(values_first1),
      values_first2(values_first2),
      keys_result(keys_result), values_result(values_result),
      comp(comp), grain_size(grain_size)
  {}

  range(range& r, ::tbb::split)
    : keys_first1(r.keys_first1), keys_last1(r.keys_last1),
      keys_first2(r.keys_first2), keys_last2(r.keys_last2),
      values_first1(r.values_first1),
      values_first2(r.values_first2),
      keys_result(r.keys_result), values_result(r.values_result),
      comp(r.comp), grain_size(r.grain_size)
  {
    // we can assume n1 and n2 are not both 0
    size_t n1 = thrust::distance(keys_first1, keys_last1);
    size_t n2 = thrust::distance(keys_first2, keys_last2);

    InputIterator1 mid1 = keys_first1;
    InputIterator2 mid2 = keys_first2;

    if (n1 > n2)
    {
      mid1 += n1 / 2;
      mid2 = thrust::lower_bound(thrust::seq, keys_first2, keys_last2, raw_reference_cast(*mid1), comp);
    }
    else
    {
      mid2 += n2 / 2;
      mid1 = thrust::upper_bound(thrust::seq, keys_first1, keys_last1, raw_reference_cast(*mid2), comp);
    }

    // set first range to [keys_first1, mid1), [keys_first2, mid2), keys_result, values_result
    r.keys_last1 = mid1;
    r.keys_last2 = mid2;

    // set second range to [mid1, keys_last1), [mid2, keys_last2), keys_result + (mid1 - keys_first1) + (mid2 - keys_first2), values_result + (mid1 - keys_first1) + (mid2 - keys_first2)
    keys_first1 = mid1;
    keys_first2 = mid2;
    values_first1 += thrust::distance(r.keys_first1, mid1);
    values_first2 += thrust::distance(r.keys_first2, mid2);
    keys_result += thrust::distance(r.keys_first1, mid1) + thrust::distance(r.keys_first2, mid2);
    values_result += thrust::distance(r.keys_first1, mid1) + thrust::distance(r.keys_first2, mid2);
  }

  bool empty(void) const
  {
    return (keys_first1 == keys_last1) && (keys_first2 == keys_last2);
  }

  bool is_divisible(void) const
  {
    return static_cast<size_t>(thrust::distance(keys_first1, keys_last1) + thrust::distance(keys_first2, keys_last2)) > grain_size;
  }
};

struct body
{
  template <typename Range>
  void operator()(Range& r) const
  {
    thrust::merge_by_key(thrust::seq,
                         r.keys_first1, r.keys_last1,
                         r.keys_first2, r.keys_last2,
                         r.values_first1,
                         r.values_first2,
                         r.keys_result,
                         r.values_result,
                         r.comp);
  }
};

} // end namespace merge_by_key_detail


template<typename DerivedPolicy,
         typename InputIterator1,
         typename InputIterator2,
         typename OutputIterator,
         typename StrictWeakOrdering>
OutputIterator merge(execution_policy<DerivedPolicy> &,
                     InputIterator1 first1,
                     InputIterator1 last1,
                     InputIterator2 first2,
                     InputIterator2 last2,
                     OutputIterator result,
                     StrictWeakOrdering comp)
{
  typedef typename merge_detail::range<InputIterator1,InputIterator2,OutputIterator,StrictWeakOrdering> Range;
  typedef          merge_detail::body                                                                   Body;
  Range range(first1, last1, first2, last2, result, comp);
  Body  body;

  ::tbb::parallel_for(range, body);

  thrust::advance(result, thrust::distance(first1, last1) + thrust::distance(first2, last2));

  return result;
} // end merge()

template <typename DerivedPolicy,
          typename InputIterator1,
          typename InputIterator2,
          typename InputIterator3,
          typename InputIterator4,
          typename OutputIterator1,
          typename OutputIterator2,
          typename StrictWeakOrdering>
thrust::pair<OutputIterator1,OutputIterator2>
  merge_by_key(execution_policy<DerivedPolicy> &,
               InputIterator1 keys_first1,
               InputIterator1 keys_last1,
               InputIterator2 keys_first2,
               InputIterator2 keys_last2,
               InputIterator3 values_first3,
               InputIterator4 values_first4,
               OutputIterator1 keys_result,
               OutputIterator2 values_result,
               StrictWeakOrdering comp)
{
  typedef typename merge_by_key_detail::range<InputIterator1,InputIterator2,InputIterator3,InputIterator4,OutputIterator1,OutputIterator2,StrictWeakOrdering> Range;
  typedef          merge_by_key_detail::body                                                                                                                  Body;

  Range range(keys_first1, keys_last1, keys_first2, keys_last2, values_first3, values_first4, keys_result, values_result, comp);
  Body  body;

  ::tbb::parallel_for(range, body);

  thrust::advance(keys_result,   thrust::distance(keys_first1, keys_last1) + thrust::distance(keys_first2, keys_last2));
  thrust::advance(values_result, thrust::distance(keys_first1, keys_last1) + thrust::distance(keys_first2, keys_last2));

  return thrust::make_pair(keys_result,values_result);
}

} // end namespace detail
} // end namespace tbb
} // end namespace system
THRUST_NAMESPACE_END